AI-Powered Web App for High-Precision Audio Stem Separation

AI-Powered Web App for High-Precision Audio Stem SeparationAI-Powered Web App for High-Precision Audio Stem Separation

The network for creativity

Join 1.25M professional creatives like you

Connect with clients, get discovered, and run your business 100% commission-free

Creatives on Contra have earned over $150M and we are just getting started

Back to feedPost

Yaseen Nasir

• Jun 17

Project Description

Developed an advanced AI-powered web application that splits any audio track into isolated vocal and instrumental stems with studio-grade precision. Built the frontend using React.js and Tailwind CSS, seamlessly integrating the industry-standard Demucs AI model to handle complex audio source separation.

Key Technical Features Implemented:

Demucs AI Integration: Leveraged Demucs' deep learning architecture to cleanly isolate vocals and background music without losing audio fidelity or introducing artifacts.

Dynamic React UI: Architected an intuitive, state-driven user interface featuring smooth file drag-and-drop areas, real-time extraction progress meters, and independent audio players for individual stems.

Efficient File Pipeline: Engineered a robust processing pipeline that securely handles high-quality audio formats, executes the separation algorithm, and delivers instantly downloadable independent WAV/MP3 files.

Full Stack Development Python React

Lynksen Digital

pro

• Jun 23

Which one would you choose? 👇 Dark mode or light mode for an operational SaaS platform? ✨

Software Development UX Design Next.js React Figma saasdashboard Product Design

16 voted

30%

37 voted

70%

53 votes

Closed

Natalia Trusova

• Jun 23

I like dark as a designer and light as a user

Shiraz Ali

• Jun 24

Built a production-ready, end-to-end agentic AI system designed to automate the entire B2B lead generation and cold outreach pipeline — from discovery to inbox.

The system uses the Google Places API with dynamic neighborhood-level scanning to find local businesses that have no website. Each lead is then deeply enriched using Firecrawl, which scrapes Yelp, Facebook, Instagram, and LinkedIn to find business owners, contact emails, and detailed business context.

From there, a custom multi-agent pipeline powered by Google Gemini Flash (with OpenRouter failover) analyzes each business, identifies specific pain points a website would solve, scores them on a 1–10 opportunity scale, and generates a hyper-personalized cold outreach email tailored to their exact situation.

Everything is tracked in real-time through a modern Next.js dashboard using Server-Sent Events (SSE), with one-click options to review, edit, regenerate in different tones (Friendly, Professional, Casual), and send emails directly via Resend.